Enterprise Vocabulary Development in Protege/OWL: Workflow and - - PowerPoint PPT Presentation

Enterprise Vocabulary Development in Protege/OWL: Workflow and Concept History Requirements

Sherri de Coronado Gilberto Fragoso

Protégé Workshop – Jul 8, 2004

Topics

• Background
• NCI Thesaurus conversion to OWL
• Requirements for Using Protégé-OWL for

NCI Thesaurus

• Progress / Pilot Testing

NCI EVS

• Services and resources addressing NCI

needs for controlled vocabulary

http://ncicb.nci.nih.gov/core/EVS

• Goal: Integration by Meaning
• Collaboration between NCI OC and NCICB

– Cancer Information Products and Systems (PDQ and Cancer.gov) – caCORE and Community portals

NCICB builds on EVS via caCORE Infrastructure

caCORE caBIO API EVS Package EVS Production Servers Thesaurus Release M etathesaurus caBIO caBIO servers caBIO Repository NCICB Portals ฀ caIm age ฀ CGAP ฀ caM OD ฀ MycaBIO Hx Release XM L/RPC RM I EVS- dependent Application s

Other caBIO Packages

caDSR caDSR server caDSR Repository

caBIO API EVS PAckage

https://ncicb.nci.nih.gov/core

NCI Thesaurus

• Public domain, open content license
• Broad coverage of cancer domain

– Neoplastic disease, Findings and Abnormalities, Anatomic Structures, Agents, Cancer-related genes, Gene products, etc.

• DL based using Apelon’s Ontylog
• 34,000+ “Concepts”

– 20 hierarchies, 19 kinds – “Roles” establish semantic relationships between Concepts – “Properties” state facts about Concept

• Concept history

NCI Thesaurus Production Environment

Production Release External Testing

NCI Thesaurus Test DTS Servers

NCI Thesaurus Editing Environm ent NCI Thesaurus W orkflow

Conflict Detection and Resolution W ork List Generation Classification Hx Validation Hx Baseline Schem a Schem a Schem a Individual Editors’ TDE

• W orkflow Client
• Editing Application
• DB Schem a
• Current NCI Baseline
• Local History

Lead Editor TDE

• W ork M anager Client
• Editing Application
• Conflict Detection/Resolution
• DB Schem a
• M aster NCI Baseline
• M aster History

Change Set W ork Assign m ent Candidate Release Hx NCI Thesaurus Production DTS Servers Hx Release

Ontylog to OWL Conversion

• Why OWL Lite for the conversion?

– To make it available in a non-proprietary form – To enable a wider audience to use it. – Current Thesaurus has fairly simple semantic constructs

Mapping the Semantics

• Kinds and Concepts modeled as Classes
• Ontylog Role becomes ObjectProperty with

Domain and Range (restrictions)

• Ontylog Property becomes

AnnotationProperty

• Some and All translated as

SomeValuesFrom and AllValuesFrom

Requirements for Using Protégé- OWL

• Concept History
• Search Capabilities
• Various Edit Actions / User Interface
• Workflow Management Functions
• Vocabulary Server (DTS or something new?)

Concept History Issues

• Certain editing actions result in retirement of

Thesaurus codes

– Merge, Split, Retirement

• Dependent applications/users require a

mechanism to retrieve data coded with Thesaurus codes that have been retired

• Tracking complex edit actions in History allows

dependent apps/users to query for replacement codes

Search Capabilities

• Must operate on various term-containing

properties, not just on class names – Good search capability critical for users and editors – Search on terms in annotation properties

• Configurable, e.g. for default settings

Edit Actions / User Interface

• Support various editing actions

– Merge – Split – Pre-retirements (by editor) – Retirement (by manager)

Split Edit Action

• Generates a new class

– History must record an association between the split and the new class

• Properties and subclasses must be reviewed and resolved

between the new and existing classes

• References to existing class must be reviewed and edited if

necessary

• Must have GUI support

"Split" GUI Panel

New Class in Tree

Merge Edit Action

• Existing class is merged into another and retired

– History must record a retirement action, and an association between the surviving and the retired class

• Properties must be copied, properties of retired class

must be recorded (AnnotationProperty), subclasses must be moved to surviving class, retired class must be re-treed

• References to retired class must be reviewed and

edited if necessary

• Must have GUI support

Merge Window

Select Surviving Class, Drop into Rightmost Pane

Swap

Retirement Actions

• Editors flag class for pre-retirement

– Review and remove/modify restrictions and subclasses – State is annotated: super and subclasses, restrictions, references – References to class eliminated – Class is re-treed to holding bin, remaining subclasses re- treed under class' parent

• Manager confirms retirement

– Class is re-treed to retirement bin – No programmatic Undo support – History records the retirement action, and associations to the class' parent classes

• GUI support for pre- and retirement

Pre-Retirement GUI

Subclasses Restrictions

Workflow Management Needs

• Worklist assignments by manager and

tracking of worklist items by editors

• Assignment of editing/review privileges
• Locking and unlocking of database (or server)

for editing

• Review and consolidation of editing changes

by manager

• Generation of reports by manager or editors

Other Workflow Needs

• Import Changesets by Manager and export

Changesets by Editor (maybe)

• Export of database “Baseline” by manager

– Development or Release baselines – Release export results in auto history export

• Configuration/constraints of environment
• Backup and Restore of database to archive

by manager

Data Handling Issues

• Changed items should be flagged for review
• Consolidation/conflict resolution step

involves accepting or rejecting changes to concepts/classes made by editors

• Class/instance deletion is restricted
• All edit actions processed in parallel for

history

Progress/ Pilot Testing

• NCI Protégé/OWL extension in progress

– NCIOWLClsesTab to support workflow/ history as shown

• Pilot to Evaluate Protégé-OWL for editing

and semantic capabilities

– 2-3 months: Kevric, NCI, Stanford, Uvic

EVS Team

EVS NCI OC – oncology, pathology, pharmacy Margaret Haber Larry Wright NCI CB – biology, operations Sherri de Coronado Gilberto Fragoso Frank Hartel Apelon, Inc. Northrop Grumman, Inc. Aspen, Inc. Kevric Corporation Jim Oberthaler SAIC Stanford Medical Informatics